Skip to contentSkip to conversation

FAILED: patch "[PATCH] parisc: Remove unnecessary barriers from spinlock.h" failed to apply to 4.14-stable tree

PATCH [email protected] 2 participants 3 messages Aug 22, 2018

Subsystems (3)
Stable branch PA-RISC architecture Locking primitives
lore
Share & tools
lore

[email protected]

author maintainer
#0 lore raw

The patch below does not apply to the 4.14-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.

thanks,

greg k-h

------------------ original commit in Linus's tree ------------------

Quoted 1 line

From 3b885ac1dc35b87a39ee176a6c7e2af9c789d8b8 Mon Sep 17 00:00:00 2001

From: John David Anglin <[email protected]>
Date: Sun, 12 Aug 2018 16:31:17 -0400
Subject: [PATCH] parisc: Remove unnecessary barriers from spinlock.h
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Now that mb() is an instruction barrier, it will slow performance if we issue unnecessary barriers.

The spinlock defines have a number of unnecessary barriers.  The __ldcw() define is both a hardware and compiler barrier.  The mb() barriers in the routines using __ldcw() serve no purpose.

The only barrier needed is the one in arch_spin_unlock().  We need to ensure all accesses are complete prior to releasing the lock.

Signed-off-by John David Anglin <[email protected]>
Cc [email protected] # 4.0+
Signed-off-by Helge Deller <[email protected]>
arch/parisc/include/asm/spinlock.h +2 -6
index 6f84b6acc86e..8a63515f03bf 100644
@@ -20,7 +20,6 @@ static inline void arch_spin_lock_flags(arch_spinlock_t *x,
 {
 	volatile unsigned int *a;
 
-	mb();
 	a = __ldcw_align(x);
 	while (__ldcw(a) == 0)
 		while (*a == 0)

@@ -30,17 +29,16 @@ static inline void arch_spin_lock_flags(arch_spinlock_t *x,
 				local_irq_disable();
 			} else
 				cpu_relax();
-	mb();
 }
 #define arch_spin_lock_flags arch_spin_lock_flags
 
 static inline void arch_spin_unlock(arch_spinlock_t *x)
 {
 	volatile unsigned int *a;
-	mb();
+
 	a = __ldcw_align(x);
-	*a = 1;
 	mb();
+	*a = 1;
 }
 
 static inline int arch_spin_trylock(arch_spinlock_t *x)

@@ -48,10 +46,8 @@ static inline int arch_spin_trylock(arch_spinlock_t *x)
 	volatile unsigned int *a;
 	int ret;
 
-	mb();
 	a = __ldcw_align(x);
         ret = __ldcw(a) != 0;
-	mb();
 
 	return ret;
 }
maintainer
#1 lore raw

* [email protected] <[email protected]>:

Quoted 4 lines

The patch below does not apply to the 4.14-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.

Upstream commit ID 3b885ac1dc35b87a39ee176a6c7e2af9c789d8b8 was tagged to be backported to v4.0+.

It applied cleanly to v4.17 and v4.18.
It did not applied to v4.14 and below (to v4.0).

Here is the backport which should apply.
Thanks!
Helge

arch/parisc/include/asm/spinlock.h +2 -6
index e32936cd7f10..7031483b110c 100644
@@ -26,7 +26,6 @@ static inline void arch_spin_lock_flags(arch_spinlock_t *x,
 {
 	volatile unsigned int *a;
 
-	mb();
 	a = __ldcw_align(x);
 	while (__ldcw(a) == 0)
 		while (*a == 0)

@@ -36,16 +35,15 @@ static inline void arch_spin_lock_flags(arch_spinlock_t *x,
 				local_irq_disable();
 			} else
 				cpu_relax();
-	mb();
 }
 
 static inline void arch_spin_unlock(arch_spinlock_t *x)
 {
 	volatile unsigned int *a;
-	mb();
+
 	a = __ldcw_align(x);
-	*a = 1;
 	mb();
+	*a = 1;
 }
 
 static inline int arch_spin_trylock(arch_spinlock_t *x)

@@ -53,10 +51,8 @@ static inline int arch_spin_trylock(arch_spinlock_t *x)
 	volatile unsigned int *a;
 	int ret;
 
-	mb();
 	a = __ldcw_align(x);
         ret = __ldcw(a) != 0;
-	mb();
 
 	return ret;
 }
Quoted 24 lines

------------------ original commit in Linus's tree ------------------

From 3b885ac1dc35b87a39ee176a6c7e2af9c789d8b8 Mon Sep 17 00:00:00 2001
From: John David Anglin <[email protected]>
Date: Sun, 12 Aug 2018 16:31:17 -0400
Subject: [PATCH] parisc: Remove unnecessary barriers from spinlock.h
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Now that mb() is an instruction barrier, it will slow performance if we issue unnecessary barriers.

The spinlock defines have a number of unnecessary barriers.ￜ The __ldcw() define is both a hardware and compiler barrier.ￜ The mb() barriers in the routines using __ldcw() serve no purpose.

The only barrier needed is the one in arch_spin_unlock().ￜ We need to ensure all accesses are complete prior to releasing the lock.

Signed-off-by John David Anglin <[email protected]>
Cc [email protected] # 4.0+
Signed-off-by Helge Deller <[email protected]>
author maintainer
#2 lore raw
Quoted 7 lines · Helge Deller

* [email protected] <[email protected]>:

Quoted 4 lines

The patch below does not apply to the 4.14-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.

Quoted 7 lines

Upstream commit ID 3b885ac1dc35b87a39ee176a6c7e2af9c789d8b8 was tagged to be backported to v4.0+.

It applied cleanly to v4.17 and v4.18.
It did not applied to v4.14 and below (to v4.0).

Here is the backport which should apply.

Thanks, now queued up.

greg k-h